[omniORB] Location forward ignored by IBM's ORB

[baileyk at schneider.com]

In case anyone is curious, here's the most likely suspect for my problem.
If there's any more IBM WebSphere users out there that can decipher the
description and explain why it's considered an "application error" I'd love
to hear it.  LOCATION_FORWARD seems like such basic CORBA functionality,
it's hard to believe another ORB can get away with not supporting it
properly, or at least give a good explanation of why it can not.

http://www-1.ibm.com/support/docview.wss?rs=0&q1=InterceptorManager&uid=swg1PQ82971&loc=en_US&cs=utf-8&cc=us&lang=en

The stack trace I'm seeing in WebSphere after omniORB replys with
LOCATION_FORWARD is identical to the one in the page referenced above.


Thanks,
Kendall



Sorry to be so cheerful about it all, but
I've been working with omniORB and the IBM ORB for a while
now, and my experience would be that if you've found an
incompatibility, there's a 90% chance that the problem lies
on the IBM side.  If you're feeling brave you may be able to
hack omniORB to work around the IBM bug.

In the end we replaced the IBM ORB with the free pure-Java
JacORB implementation.  The act of replacing the default ORB
in WebSphere was not easy, but it fixed so many bugs it was
more than worth it.

Good luck.

> I'm working on a simplified ImR-like service for load balancing,
> written in Python using omniORBpy.  Some tests have shown a WebSphere
> client ignore LOCATION_FORWARD replies and just retry on the original
> endpoint.  I'm working on getting deeper traces of the WebSphere side,
> and support from IBM, but I'll post some omniORB traces here in case
> someone can detect anything unusual.
>